Deepsoilnet: A CNN-Based Framework With Gabor And LBP Feature Fusion For Automated Soil Texture Classification From Field Images

Soil texture plays a pivotal role in determining water retention, nutrient availability, and overall soil health, making its classification essential for precision agriculture and land management. This study introduces a novel deep learning-based framework for automating soil texture classification using image processing techniques combined with Convolutional Neural Networks (CNNs). The proposed system enhances soil texture analysis by integrating Gabor filters and Local Binary Patterns (LBP) for multi-scale texture extraction, followed by CNN-based classification. Soil images are first segmented to isolate relevant regions, then divided into tiles for detailed feature extraction. Comparative analysis shows that Gabor filters outperform LBP in texture enhancement, leading to improved classification performance. The model was trained and validated on a dataset comprising soil samples from varied conditions and achieved 99% accuracy with a low loss of 0.0134, demonstrating its robustness. This hybrid approach significantly advances the automation of soil analysis and presents a scalable solution for agricultural and environmental applications.